blob: 2a9d98c641ac4926eafc8a06312625fbd6218080 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/kernel/exit.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
6
Linus Torvalds1da177e2005-04-16 15:20:36 -07007#include <linux/mm.h>
8#include <linux/slab.h>
9#include <linux/interrupt.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070010#include <linux/module.h>
Randy.Dunlapc59ede72006-01-11 12:17:46 -080011#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070012#include <linux/completion.h>
13#include <linux/personality.h>
14#include <linux/tty.h>
Kirill Korotaev6b3286e2006-12-08 02:37:56 -080015#include <linux/mnt_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070016#include <linux/key.h>
17#include <linux/security.h>
18#include <linux/cpu.h>
19#include <linux/acct.h>
Jay Lan8f0ab512006-09-30 23:28:59 -070020#include <linux/tsacct_kern.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070021#include <linux/file.h>
22#include <linux/binfmts.h>
Serge E. Hallynab516012006-10-02 02:18:06 -070023#include <linux/nsproxy.h>
Sukadev Bhattiprolu84d73782006-12-08 02:38:01 -080024#include <linux/pid_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <linux/ptrace.h>
26#include <linux/profile.h>
27#include <linux/mount.h>
28#include <linux/proc_fs.h>
Eric W. Biederman49d769d2007-05-09 02:34:33 -070029#include <linux/kthread.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070030#include <linux/mempolicy.h>
Shailabh Nagarc7572492006-07-14 00:24:40 -070031#include <linux/taskstats_kern.h>
Shailabh Nagarca74e922006-07-14 00:24:36 -070032#include <linux/delayacct.h>
Rafael J. Wysocki83144182007-07-17 04:03:35 -070033#include <linux/freezer.h>
Paul Menageb4f48b62007-10-18 23:39:33 -070034#include <linux/cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070035#include <linux/syscalls.h>
Jesper Juhl7ed20e12005-05-01 08:59:14 -070036#include <linux/signal.h>
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080037#include <linux/posix-timers.h>
Matt Helsley9f460802005-11-07 00:59:16 -080038#include <linux/cn_proc.h>
Ingo Molnarde5097c2006-01-09 15:59:21 -080039#include <linux/mutex.h>
Ingo Molnar0771dfe2006-03-27 01:16:22 -080040#include <linux/futex.h>
Ingo Molnar34f192c2006-03-27 01:16:24 -080041#include <linux/compat.h>
Jens Axboeb92ce552006-04-11 13:52:07 +020042#include <linux/pipe_fs_i.h>
Al Virofa84cb92006-03-29 20:30:19 -050043#include <linux/audit.h> /* for audit_free() */
Adrian Bunk83cc5ed2006-06-25 05:47:41 -070044#include <linux/resource.h>
David Howells0d67a462006-08-29 19:05:56 +010045#include <linux/blkdev.h>
Eric Dumazet6eaeeab2007-05-10 22:22:37 -070046#include <linux/task_io_accounting_ops.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070047
48#include <asm/uaccess.h>
49#include <asm/unistd.h>
50#include <asm/pgtable.h>
51#include <asm/mmu_context.h>
52
Adrian Bunk408b6642005-05-01 08:59:29 -070053static void exit_mm(struct task_struct * tsk);
54
Linus Torvalds1da177e2005-04-16 15:20:36 -070055static void __unhash_process(struct task_struct *p)
56{
57 nr_threads--;
58 detach_pid(p, PIDTYPE_PID);
Linus Torvalds1da177e2005-04-16 15:20:36 -070059 if (thread_group_leader(p)) {
60 detach_pid(p, PIDTYPE_PGID);
61 detach_pid(p, PIDTYPE_SID);
Oleg Nesterovc97d9892006-03-28 16:11:06 -080062
Eric W. Biederman5e85d4a2006-04-18 22:20:16 -070063 list_del_rcu(&p->tasks);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -080064 __get_cpu_var(process_counts)--;
Linus Torvalds1da177e2005-04-16 15:20:36 -070065 }
Oleg Nesterov47e65322006-03-28 16:11:25 -080066 list_del_rcu(&p->thread_group);
Oleg Nesterovc97d9892006-03-28 16:11:06 -080067 remove_parent(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -070068}
69
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080070/*
71 * This function expects the tasklist_lock write-locked.
72 */
73static void __exit_signal(struct task_struct *tsk)
74{
75 struct signal_struct *sig = tsk->signal;
76 struct sighand_struct *sighand;
77
78 BUG_ON(!sig);
79 BUG_ON(!atomic_read(&sig->count));
80
81 rcu_read_lock();
82 sighand = rcu_dereference(tsk->sighand);
83 spin_lock(&sighand->siglock);
84
85 posix_cpu_timers_exit(tsk);
86 if (atomic_dec_and_test(&sig->count))
87 posix_cpu_timers_exit_group(tsk);
88 else {
89 /*
90 * If there is any task waiting for the group exit
91 * then notify it:
92 */
Oleg Nesterov6db840f2007-10-16 23:27:23 -070093 if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count)
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080094 wake_up_process(sig->group_exit_task);
Oleg Nesterov6db840f2007-10-16 23:27:23 -070095
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080096 if (tsk == sig->curr_target)
97 sig->curr_target = next_thread(tsk);
98 /*
99 * Accumulate here the counters for all threads but the
100 * group leader as they die, so they can be added into
101 * the process-wide totals when those are taken.
102 * The group leader stays around as a zombie as long
103 * as there are other threads. When it gets reaped,
104 * the exit.c code will add its counts into these totals.
105 * We won't ever get here for the group leader, since it
106 * will have been the last reference on the signal_struct.
107 */
108 sig->utime = cputime_add(sig->utime, tsk->utime);
109 sig->stime = cputime_add(sig->stime, tsk->stime);
Laurent Vivier9ac52312007-10-15 17:00:19 +0200110 sig->gtime = cputime_add(sig->gtime, tsk->gtime);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800111 sig->min_flt += tsk->min_flt;
112 sig->maj_flt += tsk->maj_flt;
113 sig->nvcsw += tsk->nvcsw;
114 sig->nivcsw += tsk->nivcsw;
Eric Dumazet6eaeeab2007-05-10 22:22:37 -0700115 sig->inblock += task_io_get_inblock(tsk);
116 sig->oublock += task_io_get_oublock(tsk);
Balbir Singh172ba842007-07-09 18:52:00 +0200117 sig->sum_sched_runtime += tsk->se.sum_exec_runtime;
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800118 sig = NULL; /* Marker for below. */
119 }
120
Oleg Nesterov58767002006-03-28 16:11:20 -0800121 __unhash_process(tsk);
122
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800123 tsk->signal = NULL;
Oleg Nesterova7e53282006-03-28 16:11:27 -0800124 tsk->sighand = NULL;
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800125 spin_unlock(&sighand->siglock);
126 rcu_read_unlock();
127
Oleg Nesterova7e53282006-03-28 16:11:27 -0800128 __cleanup_sighand(sighand);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800129 clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
130 flush_sigqueue(&tsk->pending);
131 if (sig) {
132 flush_sigqueue(&sig->shared_pending);
Oleg Nesterov093a8e82006-10-28 10:38:51 -0700133 taskstats_tgid_free(sig);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800134 __cleanup_signal(sig);
135 }
136}
137
Eric W. Biederman8c7904a2006-03-31 02:31:37 -0800138static void delayed_put_task_struct(struct rcu_head *rhp)
139{
140 put_task_struct(container_of(rhp, struct task_struct, rcu));
141}
142
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143void release_task(struct task_struct * p)
144{
Ingo Molnar36c8b582006-07-03 00:25:41 -0700145 struct task_struct *leader;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146 int zap_leader;
Oleg Nesterov1f09f972006-03-28 16:11:11 -0800147repeat:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148 atomic_dec(&p->user->processes);
Pavel Emelyanov60347f62007-10-18 23:40:03 -0700149 proc_flush_task(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150 write_lock_irq(&tasklist_lock);
Oleg Nesterov1f09f972006-03-28 16:11:11 -0800151 ptrace_unlink(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152 BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
153 __exit_signal(p);
Oleg Nesterov35f5cad2006-03-28 16:11:19 -0800154
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155 /*
156 * If we are the last non-leader member of the thread
157 * group, and the leader is zombie, then notify the
158 * group leader's parent process. (if it wants notification.)
159 */
160 zap_leader = 0;
161 leader = p->group_leader;
162 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
163 BUG_ON(leader->exit_signal == -1);
164 do_notify_parent(leader, leader->exit_signal);
165 /*
166 * If we were the last child thread and the leader has
167 * exited already, and the leader's parent ignores SIGCHLD,
168 * then we are the one who should release the leader.
169 *
170 * do_notify_parent() will have marked it self-reaping in
171 * that case.
172 */
173 zap_leader = (leader->exit_signal == -1);
174 }
175
Linus Torvalds1da177e2005-04-16 15:20:36 -0700176 write_unlock_irq(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177 release_thread(p);
Eric W. Biederman8c7904a2006-03-31 02:31:37 -0800178 call_rcu(&p->rcu, delayed_put_task_struct);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179
180 p = leader;
181 if (unlikely(zap_leader))
182 goto repeat;
183}
184
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185/*
186 * This checks not only the pgrp, but falls back on the pid if no
187 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
188 * without this...
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800189 *
190 * The caller must hold rcu lock or the tasklist lock.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191 */
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800192struct pid *session_of_pgrp(struct pid *pgrp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193{
194 struct task_struct *p;
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800195 struct pid *sid = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700196
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800197 p = pid_task(pgrp, PIDTYPE_PGID);
Oleg Nesterov62dfb552006-12-08 02:38:03 -0800198 if (p == NULL)
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800199 p = pid_task(pgrp, PIDTYPE_PID);
Oleg Nesterov62dfb552006-12-08 02:38:03 -0800200 if (p != NULL)
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800201 sid = task_session(p);
Oleg Nesterov62dfb552006-12-08 02:38:03 -0800202
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203 return sid;
204}
205
206/*
207 * Determine if a process group is "orphaned", according to the POSIX
208 * definition in 2.2.2.52. Orphaned process groups are not to be affected
209 * by terminal-generated stop signals. Newly orphaned process groups are
210 * to receive a SIGHUP and a SIGCONT.
211 *
212 * "I ask you, have you ever known what it is to be an orphan?"
213 */
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800214static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215{
216 struct task_struct *p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800218 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300219 if ((p == ignored_task) ||
220 (p->exit_state && thread_group_empty(p)) ||
221 is_global_init(p->real_parent))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222 continue;
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300223
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800224 if (task_pgrp(p->real_parent) != pgrp &&
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300225 task_session(p->real_parent) == task_session(p))
226 return 0;
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800227 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300228
229 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230}
231
Eric W. Biederman3e7cd6c2007-02-12 00:52:58 -0800232int is_current_pgrp_orphaned(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233{
234 int retval;
235
236 read_lock(&tasklist_lock);
Eric W. Biederman3e7cd6c2007-02-12 00:52:58 -0800237 retval = will_become_orphaned_pgrp(task_pgrp(current), NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238 read_unlock(&tasklist_lock);
239
240 return retval;
241}
242
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800243static int has_stopped_jobs(struct pid *pgrp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244{
245 int retval = 0;
246 struct task_struct *p;
247
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800248 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
Matthew Wilcox338077e2007-12-06 11:09:35 -0500249 if (!task_is_stopped(p))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251 retval = 1;
252 break;
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800253 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254 return retval;
255}
256
Oleg Nesterovf49ee502008-03-02 21:44:40 +0300257/*
258 * Check to see if any process groups have become orphaned as
259 * a result of our exiting, and if they have any stopped jobs,
260 * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
261 */
262static void
263kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent)
264{
265 struct pid *pgrp = task_pgrp(tsk);
266 struct task_struct *ignored_task = tsk;
267
268 if (!parent)
269 /* exit: our father is in a different pgrp than
270 * we are and we were the only connection outside.
271 */
272 parent = tsk->real_parent;
273 else
274 /* reparent: our child is in a different pgrp than
275 * we are, and it was the only connection outside.
276 */
277 ignored_task = NULL;
278
279 if (task_pgrp(parent) != pgrp &&
280 task_session(parent) == task_session(tsk) &&
281 will_become_orphaned_pgrp(pgrp, ignored_task) &&
282 has_stopped_jobs(pgrp)) {
283 __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp);
284 __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp);
285 }
286}
287
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288/**
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700289 * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290 *
291 * If a kernel thread is launched as a result of a system call, or if
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700292 * it ever exits, it should generally reparent itself to kthreadd so it
293 * isn't in the way of other processes and is correctly cleaned up on exit.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 *
295 * The various task state such as scheduling policy and priority may have
296 * been inherited from a user process, so we reset them to sane values here.
297 *
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700298 * NOTE that reparent_to_kthreadd() gives the caller full capabilities.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299 */
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700300static void reparent_to_kthreadd(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301{
302 write_lock_irq(&tasklist_lock);
303
304 ptrace_unlink(current);
305 /* Reparent to init */
Oleg Nesterov9b678ec2006-03-28 16:11:05 -0800306 remove_parent(current);
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700307 current->real_parent = current->parent = kthreadd_task;
Oleg Nesterov9b678ec2006-03-28 16:11:05 -0800308 add_parent(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309
310 /* Set the exit signal to SIGCHLD so we signal init on exit */
311 current->exit_signal = SIGCHLD;
312
Ingo Molnare05606d2007-07-09 18:51:59 +0200313 if (task_nice(current) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314 set_user_nice(current, 0);
315 /* cpus_allowed? */
316 /* rt_priority? */
317 /* signals? */
318 security_task_reparent_to_init(current);
319 memcpy(current->signal->rlim, init_task.signal->rlim,
320 sizeof(current->signal->rlim));
321 atomic_inc(&(INIT_USER->__count));
322 write_unlock_irq(&tasklist_lock);
323 switch_uid(INIT_USER);
324}
325
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800326void __set_special_pids(struct pid *pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327{
Oren Laadane19f2472006-01-08 01:03:58 -0800328 struct task_struct *curr = current->group_leader;
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800329 pid_t nr = pid_nr(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800331 if (task_session(curr) != pid) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332 detach_pid(curr, PIDTYPE_SID);
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800333 attach_pid(curr, PIDTYPE_SID, pid);
334 set_task_session(curr, nr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335 }
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800336 if (task_pgrp(curr) != pid) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337 detach_pid(curr, PIDTYPE_PGID);
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800338 attach_pid(curr, PIDTYPE_PGID, pid);
339 set_task_pgrp(curr, nr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340 }
341}
342
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800343static void set_special_pids(struct pid *pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344{
345 write_lock_irq(&tasklist_lock);
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800346 __set_special_pids(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347 write_unlock_irq(&tasklist_lock);
348}
349
350/*
351 * Let kernel threads use this to say that they
352 * allow a certain signal (since daemonize() will
353 * have disabled all of them by default).
354 */
355int allow_signal(int sig)
356{
Jesper Juhl7ed20e12005-05-01 08:59:14 -0700357 if (!valid_signal(sig) || sig < 1)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358 return -EINVAL;
359
360 spin_lock_irq(&current->sighand->siglock);
361 sigdelset(&current->blocked, sig);
362 if (!current->mm) {
363 /* Kernel threads handle their own signals.
364 Let the signal code know it'll be handled, so
365 that they don't get converted to SIGKILL or
366 just silently dropped */
367 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
368 }
369 recalc_sigpending();
370 spin_unlock_irq(&current->sighand->siglock);
371 return 0;
372}
373
374EXPORT_SYMBOL(allow_signal);
375
376int disallow_signal(int sig)
377{
Jesper Juhl7ed20e12005-05-01 08:59:14 -0700378 if (!valid_signal(sig) || sig < 1)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379 return -EINVAL;
380
381 spin_lock_irq(&current->sighand->siglock);
Oleg Nesterov10ab8252007-05-09 02:34:37 -0700382 current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 recalc_sigpending();
384 spin_unlock_irq(&current->sighand->siglock);
385 return 0;
386}
387
388EXPORT_SYMBOL(disallow_signal);
389
390/*
391 * Put all the gunge required to become a kernel thread without
392 * attached user resources in one place where it belongs.
393 */
394
395void daemonize(const char *name, ...)
396{
397 va_list args;
398 struct fs_struct *fs;
399 sigset_t blocked;
400
401 va_start(args, name);
402 vsnprintf(current->comm, sizeof(current->comm), name, args);
403 va_end(args);
404
405 /*
406 * If we were started as result of loading a module, close all of the
407 * user space pages. We don't need them, and if we didn't close them
408 * they would be locked into memory.
409 */
410 exit_mm(current);
Rafael J. Wysocki83144182007-07-17 04:03:35 -0700411 /*
412 * We don't want to have TIF_FREEZE set if the system-wide hibernation
413 * or suspend transition begins right now.
414 */
415 current->flags |= PF_NOFREEZE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800417 if (current->nsproxy != &init_nsproxy) {
418 get_nsproxy(&init_nsproxy);
419 switch_task_namespaces(current, &init_nsproxy);
420 }
Oleg Nesterov297bd422008-02-08 04:19:10 -0800421 set_special_pids(&init_struct_pid);
Peter Zijlstra24ec8392006-12-08 02:36:04 -0800422 proc_clear_tty(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423
424 /* Block and flush all signals */
425 sigfillset(&blocked);
426 sigprocmask(SIG_BLOCK, &blocked, NULL);
427 flush_signals(current);
428
429 /* Become as one with the init task */
430
431 exit_fs(current); /* current->fs->count--; */
432 fs = init_task.fs;
433 current->fs = fs;
434 atomic_inc(&fs->count);
Serge E. Hallynab516012006-10-02 02:18:06 -0700435
Daniel Walkerd4c5e412007-10-18 23:39:59 -0700436 exit_files(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 current->files = init_task.files;
438 atomic_inc(&current->files->count);
439
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700440 reparent_to_kthreadd();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441}
442
443EXPORT_SYMBOL(daemonize);
444
Arjan van de Ven858119e2006-01-14 13:20:43 -0800445static void close_files(struct files_struct * files)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446{
447 int i, j;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700448 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449
450 j = 0;
Dipankar Sarma4fb3a532005-09-16 19:28:13 -0700451
452 /*
453 * It is safe to dereference the fd table without RCU or
454 * ->file_lock because this is the last reference to the
455 * files structure.
456 */
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700457 fdt = files_fdtable(files);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458 for (;;) {
459 unsigned long set;
460 i = j * __NFDBITS;
Vadim Lobanovbbea9f62006-12-10 02:21:12 -0800461 if (i >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462 break;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700463 set = fdt->open_fds->fds_bits[j++];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464 while (set) {
465 if (set & 1) {
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700466 struct file * file = xchg(&fdt->fd[i], NULL);
Ingo Molnar944be0b2007-02-12 00:52:26 -0800467 if (file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468 filp_close(file, files);
Ingo Molnar944be0b2007-02-12 00:52:26 -0800469 cond_resched();
470 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471 }
472 i++;
473 set >>= 1;
474 }
475 }
476}
477
478struct files_struct *get_files_struct(struct task_struct *task)
479{
480 struct files_struct *files;
481
482 task_lock(task);
483 files = task->files;
484 if (files)
485 atomic_inc(&files->count);
486 task_unlock(task);
487
488 return files;
489}
490
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800491void put_files_struct(struct files_struct *files)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492{
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700493 struct fdtable *fdt;
494
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495 if (atomic_dec_and_test(&files->count)) {
496 close_files(files);
497 /*
498 * Free the fd and fdset arrays if we expanded them.
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700499 * If the fdtable was embedded, pass files for freeing
500 * at the end of the RCU grace period. Otherwise,
501 * you can free files immediately.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502 */
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700503 fdt = files_fdtable(files);
Vadim Lobanov4fd45812006-12-10 02:21:17 -0800504 if (fdt != &files->fdtab)
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700505 kmem_cache_free(files_cachep, files);
Vadim Lobanov01b2d932006-12-22 01:10:43 -0800506 free_fdtable(fdt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507 }
508}
509
Al Viro3b125382008-04-22 05:31:30 -0400510void reset_files_struct(struct files_struct *files)
Kirill Korotaev3b9b8ab2006-09-29 02:00:05 -0700511{
Al Viro3b125382008-04-22 05:31:30 -0400512 struct task_struct *tsk = current;
Kirill Korotaev3b9b8ab2006-09-29 02:00:05 -0700513 struct files_struct *old;
514
515 old = tsk->files;
516 task_lock(tsk);
517 tsk->files = files;
518 task_unlock(tsk);
519 put_files_struct(old);
520}
Kirill Korotaev3b9b8ab2006-09-29 02:00:05 -0700521
Al Viro1ec7f1d2008-04-22 05:35:42 -0400522void exit_files(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523{
524 struct files_struct * files = tsk->files;
525
526 if (files) {
527 task_lock(tsk);
528 tsk->files = NULL;
529 task_unlock(tsk);
530 put_files_struct(files);
531 }
532}
533
Al Viro1ec7f1d2008-04-22 05:35:42 -0400534void put_fs_struct(struct fs_struct *fs)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535{
536 /* No need to hold fs->lock if we are killing it */
537 if (atomic_dec_and_test(&fs->count)) {
Jan Blunck6ac08c32008-02-14 19:34:38 -0800538 path_put(&fs->root);
539 path_put(&fs->pwd);
540 if (fs->altroot.dentry)
541 path_put(&fs->altroot);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542 kmem_cache_free(fs_cachep, fs);
543 }
544}
545
Al Viro1ec7f1d2008-04-22 05:35:42 -0400546void exit_fs(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547{
548 struct fs_struct * fs = tsk->fs;
549
550 if (fs) {
551 task_lock(tsk);
552 tsk->fs = NULL;
553 task_unlock(tsk);
Al Viro1ec7f1d2008-04-22 05:35:42 -0400554 put_fs_struct(fs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555 }
556}
557
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558EXPORT_SYMBOL_GPL(exit_fs);
559
560/*
561 * Turn us into a lazy TLB process if we
562 * aren't already..
563 */
Adrian Bunk408b6642005-05-01 08:59:29 -0700564static void exit_mm(struct task_struct * tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565{
566 struct mm_struct *mm = tsk->mm;
567
568 mm_release(tsk, mm);
569 if (!mm)
570 return;
571 /*
572 * Serialize with any possible pending coredump.
573 * We must hold mmap_sem around checking core_waiters
574 * and clearing tsk->mm. The core-inducing thread
575 * will increment core_waiters for each thread in the
576 * group with ->mm != NULL.
577 */
578 down_read(&mm->mmap_sem);
579 if (mm->core_waiters) {
580 up_read(&mm->mmap_sem);
581 down_write(&mm->mmap_sem);
582 if (!--mm->core_waiters)
583 complete(mm->core_startup_done);
584 up_write(&mm->mmap_sem);
585
586 wait_for_completion(&mm->core_done);
587 down_read(&mm->mmap_sem);
588 }
589 atomic_inc(&mm->mm_count);
Eric Sesterhenn125e1872006-06-23 02:06:06 -0700590 BUG_ON(mm != tsk->active_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591 /* more a memory barrier than a real lock */
592 task_lock(tsk);
593 tsk->mm = NULL;
594 up_read(&mm->mmap_sem);
595 enter_lazy_tlb(mm, current);
Rafael J. Wysocki0c1eecf2007-07-19 01:47:33 -0700596 /* We don't want this task to be frozen prematurely */
597 clear_freeze_flag(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598 task_unlock(tsk);
599 mmput(mm);
600}
601
Ingo Molnar36c8b582006-07-03 00:25:41 -0700602static void
603reparent_thread(struct task_struct *p, struct task_struct *father, int traced)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604{
Oleg Nesterov241ceee2006-12-24 23:30:44 +0300605 if (p->pdeath_signal)
606 /* We already hold the tasklist_lock here. */
607 group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p);
608
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609 /* Move the child from its dying parent to the new one. */
610 if (unlikely(traced)) {
611 /* Preserve ptrace links if someone else is tracing this child. */
612 list_del_init(&p->ptrace_list);
613 if (p->parent != p->real_parent)
614 list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
615 } else {
616 /* If this child is being traced, then we're the one tracing it
617 * anyway, so let go of it.
618 */
619 p->ptrace = 0;
Oleg Nesterov6ac781b2006-03-28 16:11:09 -0800620 remove_parent(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621 p->parent = p->real_parent;
Oleg Nesterov6ac781b2006-03-28 16:11:09 -0800622 add_parent(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623
Matthew Wilcox338077e2007-12-06 11:09:35 -0500624 if (task_is_traced(p)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625 /*
626 * If it was at a trace stop, turn it into
627 * a normal stop since it's no longer being
628 * traced.
629 */
630 ptrace_untrace(p);
631 }
632 }
633
Eric W. Biedermanb2b2cbc2006-12-21 21:28:40 -0700634 /* If this is a threaded reparent there is no need to
635 * notify anyone anything has happened.
636 */
637 if (p->real_parent->group_leader == father->group_leader)
638 return;
639
640 /* We don't want people slaying init. */
641 if (p->exit_signal != -1)
642 p->exit_signal = SIGCHLD;
Eric W. Biedermanb2b2cbc2006-12-21 21:28:40 -0700643
644 /* If we'd notified the old parent about this child's death,
645 * also notify the new parent.
646 */
647 if (!traced && p->exit_state == EXIT_ZOMBIE &&
648 p->exit_signal != -1 && thread_group_empty(p))
649 do_notify_parent(p, p->exit_signal);
650
Oleg Nesterovf49ee502008-03-02 21:44:40 +0300651 kill_orphaned_pgrp(p, father);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652}
653
654/*
655 * When we die, we re-parent all our children.
656 * Try to give them to another thread in our thread
657 * group, and if no such member exists, give it to
Sukadev Bhattiprolu84d73782006-12-08 02:38:01 -0800658 * the child reaper process (ie "init") in our pid
659 * space.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660 */
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700661static void forget_original_parent(struct task_struct *father)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662{
Matthias Kaehlcke03ff1792007-10-18 23:39:57 -0700663 struct task_struct *p, *n, *reaper = father;
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700664 struct list_head ptrace_dead;
665
666 INIT_LIST_HEAD(&ptrace_dead);
667
668 write_lock_irq(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669
670 do {
671 reaper = next_thread(reaper);
672 if (reaper == father) {
Sukadev Bhattiprolu88f21d82007-10-18 23:39:50 -0700673 reaper = task_child_reaper(father);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 break;
675 }
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700676 } while (reaper->flags & PF_EXITING);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677
678 /*
679 * There are only two places where our children can be:
680 *
681 * - in our child list
682 * - in our ptraced child list
683 *
684 * Search them and reparent children.
685 */
Matthias Kaehlcke03ff1792007-10-18 23:39:57 -0700686 list_for_each_entry_safe(p, n, &father->children, sibling) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687 int ptrace;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688
689 ptrace = p->ptrace;
690
691 /* if father isn't the real parent, then ptrace must be enabled */
692 BUG_ON(father != p->real_parent && !ptrace);
693
694 if (father == p->real_parent) {
695 /* reparent with a reaper, real father it's us */
Oleg Nesterov84eb6462007-10-16 23:26:49 -0700696 p->real_parent = reaper;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697 reparent_thread(p, father, 0);
698 } else {
699 /* reparent ptraced task to its real parent */
700 __ptrace_unlink (p);
701 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
702 thread_group_empty(p))
703 do_notify_parent(p, p->exit_signal);
704 }
705
706 /*
707 * if the ptraced child is a zombie with exit_signal == -1
708 * we must collect it before we exit, or it will remain
709 * zombie forever since we prevented it from self-reap itself
710 * while it was being traced by us, to be able to see it in wait4.
711 */
712 if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && p->exit_signal == -1))
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700713 list_add(&p->ptrace_list, &ptrace_dead);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700714 }
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700715
Matthias Kaehlcke03ff1792007-10-18 23:39:57 -0700716 list_for_each_entry_safe(p, n, &father->ptrace_children, ptrace_list) {
Oleg Nesterov84eb6462007-10-16 23:26:49 -0700717 p->real_parent = reaper;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 reparent_thread(p, father, 1);
719 }
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700720
721 write_unlock_irq(&tasklist_lock);
722 BUG_ON(!list_empty(&father->children));
723 BUG_ON(!list_empty(&father->ptrace_children));
724
725 list_for_each_entry_safe(p, n, &ptrace_dead, ptrace_list) {
726 list_del_init(&p->ptrace_list);
727 release_task(p);
728 }
729
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730}
731
732/*
733 * Send signals to all our closest relatives so that they know
734 * to properly mourn us..
735 */
Oleg Nesterov821c7de2008-03-02 21:44:44 +0300736static void exit_notify(struct task_struct *tsk, int group_dead)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737{
738 int state;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740 /*
741 * This does two things:
742 *
743 * A. Make init inherit all the child processes
744 * B. Check to see if any process groups have become orphaned
745 * as a result of our exiting, and if they have any stopped
746 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
747 */
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700748 forget_original_parent(tsk);
Pavel Emelyanov2e4a7072007-10-18 23:40:01 -0700749 exit_task_namespaces(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700751 write_lock_irq(&tasklist_lock);
Oleg Nesterov821c7de2008-03-02 21:44:44 +0300752 if (group_dead)
753 kill_orphaned_pgrp(tsk->group_leader, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754
Oleg Nesterov24728442007-08-04 01:04:41 +0400755 /* Let father know we died
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756 *
757 * Thread signals are configurable, but you aren't going to use
Daniel Walkerd4c5e412007-10-18 23:39:59 -0700758 * that to send signals to arbitary processes.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759 * That stops right now.
760 *
761 * If the parent exec id doesn't match the exec id we saved
762 * when we started then we know the parent has changed security
763 * domain.
764 *
765 * If our self_exec id doesn't match our parent_exec_id then
766 * we have changed execution domain as these two values started
767 * the same after a fork.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769 if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
Oleg Nesterovf49ee502008-03-02 21:44:40 +0300770 (tsk->parent_exec_id != tsk->real_parent->self_exec_id ||
771 tsk->self_exec_id != tsk->parent_exec_id)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700772 && !capable(CAP_KILL))
773 tsk->exit_signal = SIGCHLD;
774
775
776 /* If something other than our normal parent is ptracing us, then
777 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
778 * only has special meaning to our real parent.
779 */
780 if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
781 int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
782 do_notify_parent(tsk, signal);
783 } else if (tsk->ptrace) {
784 do_notify_parent(tsk, SIGCHLD);
785 }
786
787 state = EXIT_ZOMBIE;
Oleg Nesterov24728442007-08-04 01:04:41 +0400788 if (tsk->exit_signal == -1 && likely(!tsk->ptrace))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789 state = EXIT_DEAD;
790 tsk->exit_state = state;
791
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700792 if (thread_group_leader(tsk) &&
793 tsk->signal->notify_count < 0 &&
794 tsk->signal->group_exit_task)
795 wake_up_process(tsk->signal->group_exit_task);
796
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797 write_unlock_irq(&tasklist_lock);
798
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799 /* If the process is dead, release it - nobody will wait for it */
800 if (state == EXIT_DEAD)
801 release_task(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802}
803
Jeff Dikee18eecb2007-07-15 23:38:48 -0700804#ifdef CONFIG_DEBUG_STACK_USAGE
805static void check_stack_usage(void)
806{
807 static DEFINE_SPINLOCK(low_water_lock);
808 static int lowest_to_date = THREAD_SIZE;
809 unsigned long *n = end_of_stack(current);
810 unsigned long free;
811
812 while (*n == 0)
813 n++;
814 free = (unsigned long)n - (unsigned long)end_of_stack(current);
815
816 if (free >= lowest_to_date)
817 return;
818
819 spin_lock(&low_water_lock);
820 if (free < lowest_to_date) {
821 printk(KERN_WARNING "%s used greatest stack depth: %lu bytes "
822 "left\n",
823 current->comm, free);
824 lowest_to_date = free;
825 }
826 spin_unlock(&low_water_lock);
827}
828#else
829static inline void check_stack_usage(void) {}
830#endif
831
Oleg Nesterov84eb6462007-10-16 23:26:49 -0700832static inline void exit_child_reaper(struct task_struct *tsk)
833{
Sukadev Bhattiprolu88f21d82007-10-18 23:39:50 -0700834 if (likely(tsk->group_leader != task_child_reaper(tsk)))
Oleg Nesterov84eb6462007-10-16 23:26:49 -0700835 return;
836
Sukadev Bhattiprolu3eb07c82007-10-18 23:40:13 -0700837 if (tsk->nsproxy->pid_ns == &init_pid_ns)
838 panic("Attempted to kill init!");
839
840 /*
841 * @tsk is the last thread in the 'cgroup-init' and is exiting.
842 * Terminate all remaining processes in the namespace and reap them
843 * before exiting @tsk.
844 *
845 * Note that @tsk (last thread of cgroup-init) may not necessarily
846 * be the child-reaper (i.e main thread of cgroup-init) of the
847 * namespace i.e the child_reaper may have already exited.
848 *
849 * Even after a child_reaper exits, we let it inherit orphaned children,
850 * because, pid_ns->child_reaper remains valid as long as there is
851 * at least one living sub-thread in the cgroup init.
852
853 * This living sub-thread of the cgroup-init will be notified when
854 * a child inherited by the 'child-reaper' exits (do_notify_parent()
855 * uses __group_send_sig_info()). Further, when reaping child processes,
856 * do_wait() iterates over children of all living sub threads.
857
858 * i.e even though 'child_reaper' thread is listed as the parent of the
859 * orphaned children, any living sub-thread in the cgroup-init can
860 * perform the role of the child_reaper.
861 */
862 zap_pid_ns_processes(tsk->nsproxy->pid_ns);
Oleg Nesterov84eb6462007-10-16 23:26:49 -0700863}
864
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800865NORET_TYPE void do_exit(long code)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866{
867 struct task_struct *tsk = current;
868 int group_dead;
869
870 profile_task_exit(tsk);
871
Jens Axboe22e2c502005-06-27 10:55:12 +0200872 WARN_ON(atomic_read(&tsk->fs_excl));
873
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874 if (unlikely(in_interrupt()))
875 panic("Aiee, killing interrupt handler!");
876 if (unlikely(!tsk->pid))
877 panic("Attempted to kill the idle task!");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878
879 if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
880 current->ptrace_message = code;
881 ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
882 }
883
Alexander Nybergdf164db2005-06-23 00:09:13 -0700884 /*
885 * We're taking recursive faults here in do_exit. Safest is to just
886 * leave this task alone and wait for reboot.
887 */
888 if (unlikely(tsk->flags & PF_EXITING)) {
889 printk(KERN_ALERT
890 "Fixing recursive fault but reboot is needed!\n");
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -0700891 /*
892 * We can do this unlocked here. The futex code uses
893 * this flag just to verify whether the pi state
894 * cleanup has been done or not. In the worst case it
895 * loops once more. We pretend that the cleanup was
896 * done as there is no way to return. Either the
897 * OWNER_DIED bit is set by now or we push the blocked
898 * task into the wait for ever nirwana as well.
899 */
900 tsk->flags |= PF_EXITPIDONE;
Al Viroafc847b2006-02-28 12:51:55 -0500901 if (tsk->io_context)
902 exit_io_context();
Alexander Nybergdf164db2005-06-23 00:09:13 -0700903 set_current_state(TASK_UNINTERRUPTIBLE);
904 schedule();
905 }
906
Oleg Nesterovd12619b2008-02-08 04:19:12 -0800907 exit_signals(tsk); /* sets PF_EXITING */
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -0700908 /*
909 * tsk->flags are checked in the futex code to protect against
910 * an exiting task cleaning up the robust pi futexes.
911 */
Oleg Nesterovd2ee7192007-10-16 23:26:47 -0700912 smp_mb();
913 spin_unlock_wait(&tsk->pi_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915 if (unlikely(in_atomic()))
916 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
Pavel Emelyanovba25f9d2007-10-18 23:40:40 -0700917 current->comm, task_pid_nr(current),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918 preempt_count());
919
920 acct_update_integrals(tsk);
Hugh Dickins365e9c872005-10-29 18:16:18 -0700921 if (tsk->mm) {
922 update_hiwater_rss(tsk->mm);
923 update_hiwater_vm(tsk->mm);
924 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925 group_dead = atomic_dec_and_test(&tsk->signal->live);
Andrew Mortonc3068952005-08-04 16:49:32 -0700926 if (group_dead) {
Oleg Nesterov84eb6462007-10-16 23:26:49 -0700927 exit_child_reaper(tsk);
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -0700928 hrtimer_cancel(&tsk->signal->real_timer);
Roland McGrath25f407f2005-10-21 15:03:29 -0700929 exit_itimers(tsk->signal);
Andrew Mortonc3068952005-08-04 16:49:32 -0700930 }
KaiGai Koheif6ec29a2006-06-25 05:49:25 -0700931 acct_collect(code, group_dead);
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -0700932#ifdef CONFIG_FUTEX
Ingo Molnar0771dfe2006-03-27 01:16:22 -0800933 if (unlikely(tsk->robust_list))
934 exit_robust_list(tsk);
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -0700935#ifdef CONFIG_COMPAT
Ingo Molnar34f192c2006-03-27 01:16:24 -0800936 if (unlikely(tsk->compat_robust_list))
937 compat_exit_robust_list(tsk);
938#endif
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -0700939#endif
Miloslav Trmac522ed772007-07-15 23:40:56 -0700940 if (group_dead)
941 tty_audit_exit();
Al Virofa84cb92006-03-29 20:30:19 -0500942 if (unlikely(tsk->audit_context))
943 audit_free(tsk);
Oleg Nesterov115085e2006-12-06 20:36:51 -0800944
Jonathan Limf2ab6d82007-08-30 23:56:23 -0700945 tsk->exit_code = code;
Oleg Nesterov115085e2006-12-06 20:36:51 -0800946 taskstats_exit(tsk, group_dead);
Shailabh Nagarc7572492006-07-14 00:24:40 -0700947
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948 exit_mm(tsk);
949
KaiGai Kohei0e464812006-06-25 05:49:24 -0700950 if (group_dead)
KaiGai Koheif6ec29a2006-06-25 05:49:25 -0700951 acct_process();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952 exit_sem(tsk);
Al Viro1ec7f1d2008-04-22 05:35:42 -0400953 exit_files(tsk);
954 exit_fs(tsk);
Jeff Dikee18eecb2007-07-15 23:38:48 -0700955 check_stack_usage();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700956 exit_thread();
Paul Menageb4f48b62007-10-18 23:39:33 -0700957 cgroup_exit(tsk, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958 exit_keys(tsk);
959
960 if (group_dead && tsk->signal->leader)
961 disassociate_ctty(1);
962
Al Viroa1261f542005-11-13 16:06:55 -0800963 module_put(task_thread_info(tsk)->exec_domain->module);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964 if (tsk->binfmt)
965 module_put(tsk->binfmt->module);
966
Matt Helsley9f460802005-11-07 00:59:16 -0800967 proc_exit_connector(tsk);
Oleg Nesterov821c7de2008-03-02 21:44:44 +0300968 exit_notify(tsk, group_dead);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969#ifdef CONFIG_NUMA
Lee Schermerhornf0be3d32008-04-28 02:13:08 -0700970 mpol_put(tsk->mempolicy);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971 tsk->mempolicy = NULL;
972#endif
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -0700973#ifdef CONFIG_FUTEX
Ingo Molnarde5097c2006-01-09 15:59:21 -0800974 /*
Ingo Molnarc87e2832006-06-27 02:54:58 -0700975 * This must happen late, after the PID is not
976 * hashed anymore:
977 */
978 if (unlikely(!list_empty(&tsk->pi_state_list)))
979 exit_pi_state_list(tsk);
980 if (unlikely(current->pi_state_cache))
981 kfree(current->pi_state_cache);
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -0700982#endif
Ingo Molnarc87e2832006-06-27 02:54:58 -0700983 /*
Ingo Molnar9a11b49a2006-07-03 00:24:33 -0700984 * Make sure we are holding no locks:
Ingo Molnarde5097c2006-01-09 15:59:21 -0800985 */
Ingo Molnar9a11b49a2006-07-03 00:24:33 -0700986 debug_check_no_locks_held(tsk);
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -0700987 /*
988 * We can do this unlocked here. The futex code uses this flag
989 * just to verify whether the pi state cleanup has been done
990 * or not. In the worst case it loops once more.
991 */
992 tsk->flags |= PF_EXITPIDONE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993
Al Viroafc847b2006-02-28 12:51:55 -0500994 if (tsk->io_context)
995 exit_io_context();
996
Jens Axboeb92ce552006-04-11 13:52:07 +0200997 if (tsk->splice_pipe)
998 __free_pipe_info(tsk->splice_pipe);
999
Coywolf Qi Hunt74072512005-10-30 15:02:47 -08001000 preempt_disable();
Oleg Nesterov55a101f2006-09-29 02:01:10 -07001001 /* causes final put_task_struct in finish_task_switch(). */
Oleg Nesterovc394cc92006-09-29 02:01:11 -07001002 tsk->state = TASK_DEAD;
Coywolf Qi Hunt74072512005-10-30 15:02:47 -08001003
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004 schedule();
1005 BUG();
1006 /* Avoid "noreturn function does return". */
Alan Cox54306cf2006-09-29 02:00:42 -07001007 for (;;)
1008 cpu_relax(); /* For when BUG is null */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009}
1010
Russ Anderson012914d2005-04-23 00:08:00 -07001011EXPORT_SYMBOL_GPL(do_exit);
1012
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013NORET_TYPE void complete_and_exit(struct completion *comp, long code)
1014{
1015 if (comp)
1016 complete(comp);
Oleg Nesterov55a101f2006-09-29 02:01:10 -07001017
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018 do_exit(code);
1019}
1020
1021EXPORT_SYMBOL(complete_and_exit);
1022
1023asmlinkage long sys_exit(int error_code)
1024{
1025 do_exit((error_code&0xff)<<8);
1026}
1027
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028/*
1029 * Take down every thread in the group. This is called by fatal signals
1030 * as well as by sys_exit_group (below).
1031 */
1032NORET_TYPE void
1033do_group_exit(int exit_code)
1034{
1035 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
1036
1037 if (current->signal->flags & SIGNAL_GROUP_EXIT)
1038 exit_code = current->signal->group_exit_code;
1039 else if (!thread_group_empty(current)) {
1040 struct signal_struct *const sig = current->signal;
1041 struct sighand_struct *const sighand = current->sighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042 spin_lock_irq(&sighand->siglock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001043 if (signal_group_exit(sig))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044 /* Another thread got here before we took the lock. */
1045 exit_code = sig->group_exit_code;
1046 else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047 sig->group_exit_code = exit_code;
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001048 sig->flags = SIGNAL_GROUP_EXIT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049 zap_other_threads(current);
1050 }
1051 spin_unlock_irq(&sighand->siglock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052 }
1053
1054 do_exit(exit_code);
1055 /* NOTREACHED */
1056}
1057
1058/*
1059 * this kills every thread in the thread group. Note that any externally
1060 * wait4()-ing process will get the correct exit code - even if this
1061 * thread is not the thread group leader.
1062 */
1063asmlinkage void sys_exit_group(int error_code)
1064{
1065 do_group_exit((error_code & 0xff) << 8);
1066}
1067
Eric W. Biederman161550d2008-02-08 04:19:14 -08001068static struct pid *task_pid_type(struct task_struct *task, enum pid_type type)
1069{
1070 struct pid *pid = NULL;
1071 if (type == PIDTYPE_PID)
1072 pid = task->pids[type].pid;
1073 else if (type < PIDTYPE_MAX)
1074 pid = task->group_leader->pids[type].pid;
1075 return pid;
1076}
1077
1078static int eligible_child(enum pid_type type, struct pid *pid, int options,
1079 struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080{
Roland McGrath73243282007-05-06 14:50:20 -07001081 int err;
1082
Eric W. Biederman161550d2008-02-08 04:19:14 -08001083 if (type < PIDTYPE_MAX) {
1084 if (task_pid_type(p, type) != pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001085 return 0;
1086 }
1087
1088 /*
1089 * Do not consider detached threads that are
1090 * not ptraced:
1091 */
1092 if (p->exit_signal == -1 && !p->ptrace)
1093 return 0;
1094
1095 /* Wait for all children (clone and not) if __WALL is set;
1096 * otherwise, wait for clone children *only* if __WCLONE is
1097 * set; otherwise, wait for non-clone children *only*. (Note:
1098 * A "clone" child here is one that reports to its parent
1099 * using a signal other than SIGCHLD.) */
1100 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
1101 && !(options & __WALL))
1102 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103
Roland McGrath73243282007-05-06 14:50:20 -07001104 err = security_task_wait(p);
Oleg Nesterovf2cc3eb2008-02-08 04:19:06 -08001105 if (likely(!err))
1106 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107
Eric W. Biederman161550d2008-02-08 04:19:14 -08001108 if (type != PIDTYPE_PID)
Oleg Nesterovf2cc3eb2008-02-08 04:19:06 -08001109 return 0;
1110 /* This child was explicitly requested, abort */
1111 read_unlock(&tasklist_lock);
1112 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001113}
1114
Ingo Molnar36c8b582006-07-03 00:25:41 -07001115static int wait_noreap_copyout(struct task_struct *p, pid_t pid, uid_t uid,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116 int why, int status,
1117 struct siginfo __user *infop,
1118 struct rusage __user *rusagep)
1119{
1120 int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
Ingo Molnar36c8b582006-07-03 00:25:41 -07001121
Linus Torvalds1da177e2005-04-16 15:20:36 -07001122 put_task_struct(p);
1123 if (!retval)
1124 retval = put_user(SIGCHLD, &infop->si_signo);
1125 if (!retval)
1126 retval = put_user(0, &infop->si_errno);
1127 if (!retval)
1128 retval = put_user((short)why, &infop->si_code);
1129 if (!retval)
1130 retval = put_user(pid, &infop->si_pid);
1131 if (!retval)
1132 retval = put_user(uid, &infop->si_uid);
1133 if (!retval)
1134 retval = put_user(status, &infop->si_status);
1135 if (!retval)
1136 retval = pid;
1137 return retval;
1138}
1139
1140/*
1141 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1142 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1143 * the lock and this task is uninteresting. If we return nonzero, we have
1144 * released the lock and the system call should return.
1145 */
Ingo Molnar36c8b582006-07-03 00:25:41 -07001146static int wait_task_zombie(struct task_struct *p, int noreap,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147 struct siginfo __user *infop,
1148 int __user *stat_addr, struct rusage __user *ru)
1149{
1150 unsigned long state;
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001151 int retval, status, traced;
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001152 pid_t pid = task_pid_vnr(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153
1154 if (unlikely(noreap)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001155 uid_t uid = p->uid;
1156 int exit_code = p->exit_code;
1157 int why, status;
1158
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159 get_task_struct(p);
1160 read_unlock(&tasklist_lock);
1161 if ((exit_code & 0x7f) == 0) {
1162 why = CLD_EXITED;
1163 status = exit_code >> 8;
1164 } else {
1165 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1166 status = exit_code & 0x7f;
1167 }
1168 return wait_noreap_copyout(p, pid, uid, why,
1169 status, infop, ru);
1170 }
1171
1172 /*
1173 * Try to move the task's state to DEAD
1174 * only one thread is allowed to do this:
1175 */
1176 state = xchg(&p->exit_state, EXIT_DEAD);
1177 if (state != EXIT_ZOMBIE) {
1178 BUG_ON(state != EXIT_DEAD);
1179 return 0;
1180 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001182 /* traced means p->ptrace, but not vice versa */
1183 traced = (p->real_parent != p->parent);
1184
1185 if (likely(!traced)) {
Jesper Juhl3795e162006-01-09 20:54:39 -08001186 struct signal_struct *psig;
1187 struct signal_struct *sig;
1188
Linus Torvalds1da177e2005-04-16 15:20:36 -07001189 /*
1190 * The resource counters for the group leader are in its
1191 * own task_struct. Those for dead threads in the group
1192 * are in its signal_struct, as are those for the child
1193 * processes it has previously reaped. All these
1194 * accumulate in the parent's signal_struct c* fields.
1195 *
1196 * We don't bother to take a lock here to protect these
1197 * p->signal fields, because they are only touched by
1198 * __exit_signal, which runs with tasklist_lock
1199 * write-locked anyway, and so is excluded here. We do
1200 * need to protect the access to p->parent->signal fields,
1201 * as other threads in the parent group can be right
1202 * here reaping other children at the same time.
1203 */
1204 spin_lock_irq(&p->parent->sighand->siglock);
Jesper Juhl3795e162006-01-09 20:54:39 -08001205 psig = p->parent->signal;
1206 sig = p->signal;
1207 psig->cutime =
1208 cputime_add(psig->cutime,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209 cputime_add(p->utime,
Jesper Juhl3795e162006-01-09 20:54:39 -08001210 cputime_add(sig->utime,
1211 sig->cutime)));
1212 psig->cstime =
1213 cputime_add(psig->cstime,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001214 cputime_add(p->stime,
Jesper Juhl3795e162006-01-09 20:54:39 -08001215 cputime_add(sig->stime,
1216 sig->cstime)));
Laurent Vivier9ac52312007-10-15 17:00:19 +02001217 psig->cgtime =
1218 cputime_add(psig->cgtime,
1219 cputime_add(p->gtime,
1220 cputime_add(sig->gtime,
1221 sig->cgtime)));
Jesper Juhl3795e162006-01-09 20:54:39 -08001222 psig->cmin_flt +=
1223 p->min_flt + sig->min_flt + sig->cmin_flt;
1224 psig->cmaj_flt +=
1225 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1226 psig->cnvcsw +=
1227 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1228 psig->cnivcsw +=
1229 p->nivcsw + sig->nivcsw + sig->cnivcsw;
Eric Dumazet6eaeeab2007-05-10 22:22:37 -07001230 psig->cinblock +=
1231 task_io_get_inblock(p) +
1232 sig->inblock + sig->cinblock;
1233 psig->coublock +=
1234 task_io_get_oublock(p) +
1235 sig->oublock + sig->coublock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001236 spin_unlock_irq(&p->parent->sighand->siglock);
1237 }
1238
1239 /*
1240 * Now we are sure this task is interesting, and no other
1241 * thread can reap it because we set its state to EXIT_DEAD.
1242 */
1243 read_unlock(&tasklist_lock);
1244
1245 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1246 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1247 ? p->signal->group_exit_code : p->exit_code;
1248 if (!retval && stat_addr)
1249 retval = put_user(status, stat_addr);
1250 if (!retval && infop)
1251 retval = put_user(SIGCHLD, &infop->si_signo);
1252 if (!retval && infop)
1253 retval = put_user(0, &infop->si_errno);
1254 if (!retval && infop) {
1255 int why;
1256
1257 if ((status & 0x7f) == 0) {
1258 why = CLD_EXITED;
1259 status >>= 8;
1260 } else {
1261 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1262 status &= 0x7f;
1263 }
1264 retval = put_user((short)why, &infop->si_code);
1265 if (!retval)
1266 retval = put_user(status, &infop->si_status);
1267 }
1268 if (!retval && infop)
Oleg Nesterov3a515e42008-02-08 04:19:07 -08001269 retval = put_user(pid, &infop->si_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001270 if (!retval && infop)
1271 retval = put_user(p->uid, &infop->si_uid);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001272 if (!retval)
Oleg Nesterov3a515e42008-02-08 04:19:07 -08001273 retval = pid;
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001274
1275 if (traced) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 write_lock_irq(&tasklist_lock);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001277 /* We dropped tasklist, ptracer could die and untrace */
1278 ptrace_unlink(p);
1279 /*
1280 * If this is not a detached task, notify the parent.
1281 * If it's still not detached after that, don't release
1282 * it now.
1283 */
1284 if (p->exit_signal != -1) {
1285 do_notify_parent(p, p->exit_signal);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001286 if (p->exit_signal != -1) {
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001287 p->exit_state = EXIT_ZOMBIE;
1288 p = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289 }
1290 }
1291 write_unlock_irq(&tasklist_lock);
1292 }
1293 if (p != NULL)
1294 release_task(p);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001295
Linus Torvalds1da177e2005-04-16 15:20:36 -07001296 return retval;
1297}
1298
1299/*
1300 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1301 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1302 * the lock and this task is uninteresting. If we return nonzero, we have
1303 * released the lock and the system call should return.
1304 */
Oleg Nesterov1bad95c2008-02-08 04:19:03 -08001305static int wait_task_stopped(struct task_struct *p,
Ingo Molnar36c8b582006-07-03 00:25:41 -07001306 int noreap, struct siginfo __user *infop,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307 int __user *stat_addr, struct rusage __user *ru)
1308{
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001309 int retval, exit_code, why;
1310 uid_t uid = 0; /* unneeded, required by compiler */
Oleg Nesterovc8950782007-11-28 16:21:24 -08001311 pid_t pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001313 exit_code = 0;
1314 spin_lock_irq(&p->sighand->siglock);
1315
1316 if (unlikely(!task_is_stopped_or_traced(p)))
1317 goto unlock_sig;
1318
Oleg Nesterov1bad95c2008-02-08 04:19:03 -08001319 if (!(p->ptrace & PT_PTRACED) && p->signal->group_stop_count > 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001320 /*
1321 * A group stop is in progress and this is the group leader.
1322 * We won't report until all threads have stopped.
1323 */
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001324 goto unlock_sig;
1325
1326 exit_code = p->exit_code;
1327 if (!exit_code)
1328 goto unlock_sig;
1329
1330 if (!noreap)
1331 p->exit_code = 0;
1332
1333 uid = p->uid;
1334unlock_sig:
1335 spin_unlock_irq(&p->sighand->siglock);
1336 if (!exit_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337 return 0;
1338
1339 /*
1340 * Now we are pretty sure this task is interesting.
1341 * Make sure it doesn't get reaped out from under us while we
1342 * give up the lock and then examine it below. We don't want to
1343 * keep holding onto the tasklist_lock while we call getrusage and
1344 * possibly take page faults for user memory.
1345 */
1346 get_task_struct(p);
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001347 pid = task_pid_vnr(p);
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001348 why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349 read_unlock(&tasklist_lock);
1350
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001351 if (unlikely(noreap))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352 return wait_noreap_copyout(p, pid, uid,
Scott James Remnante6ceb322007-11-28 16:22:07 -08001353 why, exit_code,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354 infop, ru);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001355
1356 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1357 if (!retval && stat_addr)
1358 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1359 if (!retval && infop)
1360 retval = put_user(SIGCHLD, &infop->si_signo);
1361 if (!retval && infop)
1362 retval = put_user(0, &infop->si_errno);
1363 if (!retval && infop)
Roland McGrath6efcae462008-03-08 11:41:22 -08001364 retval = put_user((short)why, &infop->si_code);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365 if (!retval && infop)
1366 retval = put_user(exit_code, &infop->si_status);
1367 if (!retval && infop)
Oleg Nesterovc8950782007-11-28 16:21:24 -08001368 retval = put_user(pid, &infop->si_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369 if (!retval && infop)
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001370 retval = put_user(uid, &infop->si_uid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001371 if (!retval)
Oleg Nesterovc8950782007-11-28 16:21:24 -08001372 retval = pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001373 put_task_struct(p);
1374
1375 BUG_ON(!retval);
1376 return retval;
1377}
1378
1379/*
1380 * Handle do_wait work for one task in a live, non-stopped state.
1381 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1382 * the lock and this task is uninteresting. If we return nonzero, we have
1383 * released the lock and the system call should return.
1384 */
Ingo Molnar36c8b582006-07-03 00:25:41 -07001385static int wait_task_continued(struct task_struct *p, int noreap,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386 struct siginfo __user *infop,
1387 int __user *stat_addr, struct rusage __user *ru)
1388{
1389 int retval;
1390 pid_t pid;
1391 uid_t uid;
1392
Linus Torvalds1da177e2005-04-16 15:20:36 -07001393 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1394 return 0;
1395
1396 spin_lock_irq(&p->sighand->siglock);
1397 /* Re-check with the lock held. */
1398 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1399 spin_unlock_irq(&p->sighand->siglock);
1400 return 0;
1401 }
1402 if (!noreap)
1403 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1404 spin_unlock_irq(&p->sighand->siglock);
1405
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001406 pid = task_pid_vnr(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407 uid = p->uid;
1408 get_task_struct(p);
1409 read_unlock(&tasklist_lock);
1410
1411 if (!infop) {
1412 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1413 put_task_struct(p);
1414 if (!retval && stat_addr)
1415 retval = put_user(0xffff, stat_addr);
1416 if (!retval)
Oleg Nesterov3a515e42008-02-08 04:19:07 -08001417 retval = pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418 } else {
1419 retval = wait_noreap_copyout(p, pid, uid,
1420 CLD_CONTINUED, SIGCONT,
1421 infop, ru);
1422 BUG_ON(retval == 0);
1423 }
1424
1425 return retval;
1426}
1427
Eric W. Biederman161550d2008-02-08 04:19:14 -08001428static long do_wait(enum pid_type type, struct pid *pid, int options,
1429 struct siginfo __user *infop, int __user *stat_addr,
1430 struct rusage __user *ru)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431{
1432 DECLARE_WAITQUEUE(wait, current);
1433 struct task_struct *tsk;
1434 int flag, retval;
1435
1436 add_wait_queue(&current->signal->wait_chldexit,&wait);
1437repeat:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001438 /* If there is nothing that can match our critier just get out */
1439 retval = -ECHILD;
1440 if ((type < PIDTYPE_MAX) && (!pid || hlist_empty(&pid->tasks[type])))
1441 goto end;
1442
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443 /*
1444 * We will set this flag if we see any child that might later
1445 * match our criteria, even if we are not able to reap it yet.
1446 */
Oleg Nesterovf2cc3eb2008-02-08 04:19:06 -08001447 flag = retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448 current->state = TASK_INTERRUPTIBLE;
1449 read_lock(&tasklist_lock);
1450 tsk = current;
1451 do {
1452 struct task_struct *p;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453
Matthias Kaehlcke03ff1792007-10-18 23:39:57 -07001454 list_for_each_entry(p, &tsk->children, sibling) {
Eric W. Biederman161550d2008-02-08 04:19:14 -08001455 int ret = eligible_child(type, pid, options, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001456 if (!ret)
1457 continue;
1458
Roland McGrath73243282007-05-06 14:50:20 -07001459 if (unlikely(ret < 0)) {
Oleg Nesterovf2cc3eb2008-02-08 04:19:06 -08001460 retval = ret;
1461 } else if (task_is_stopped_or_traced(p)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462 /*
1463 * It's stopped now, so it might later
1464 * continue, exit, or stop again.
1465 */
1466 flag = 1;
Oleg Nesterov34a17382008-02-08 04:18:59 -08001467 if (!(p->ptrace & PT_PTRACED) &&
1468 !(options & WUNTRACED))
1469 continue;
Matthew Wilcox338077e2007-12-06 11:09:35 -05001470
Oleg Nesterov1bad95c2008-02-08 04:19:03 -08001471 retval = wait_task_stopped(p,
Matthew Wilcox338077e2007-12-06 11:09:35 -05001472 (options & WNOWAIT), infop,
1473 stat_addr, ru);
Oleg Nesterov96fabbf2008-02-08 04:19:04 -08001474 } else if (p->exit_state == EXIT_ZOMBIE &&
1475 !delay_group_leader(p)) {
Matthew Wilcox338077e2007-12-06 11:09:35 -05001476 /*
Oleg Nesterov96fabbf2008-02-08 04:19:04 -08001477 * We don't reap group leaders with subthreads.
Matthew Wilcox338077e2007-12-06 11:09:35 -05001478 */
Matthew Wilcox338077e2007-12-06 11:09:35 -05001479 if (!likely(options & WEXITED))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480 continue;
Matthew Wilcox338077e2007-12-06 11:09:35 -05001481 retval = wait_task_zombie(p,
1482 (options & WNOWAIT), infop,
1483 stat_addr, ru);
Oleg Nesterov0a76fe82008-02-06 01:37:06 -08001484 } else if (p->exit_state != EXIT_DEAD) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001485 /*
1486 * It's running now, so it might later
1487 * exit, stop, or stop and then continue.
1488 */
1489 flag = 1;
1490 if (!unlikely(options & WCONTINUED))
1491 continue;
Matthew Wilcox338077e2007-12-06 11:09:35 -05001492 retval = wait_task_continued(p,
1493 (options & WNOWAIT), infop,
1494 stat_addr, ru);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001495 }
Oleg Nesterov9cbab812008-02-08 04:19:02 -08001496 if (retval != 0) /* tasklist_lock released */
1497 goto end;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001498 }
1499 if (!flag) {
Matthias Kaehlcke03ff1792007-10-18 23:39:57 -07001500 list_for_each_entry(p, &tsk->ptrace_children,
Oleg Nesterovf2cc3eb2008-02-08 04:19:06 -08001501 ptrace_list) {
Eric W. Biederman161550d2008-02-08 04:19:14 -08001502 flag = eligible_child(type, pid, options, p);
Oleg Nesterovf2cc3eb2008-02-08 04:19:06 -08001503 if (!flag)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504 continue;
Oleg Nesterovf2cc3eb2008-02-08 04:19:06 -08001505 if (likely(flag > 0))
1506 break;
1507 retval = flag;
1508 goto end;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509 }
1510 }
1511 if (options & __WNOTHREAD)
1512 break;
1513 tsk = next_thread(tsk);
Eric Sesterhenn125e1872006-06-23 02:06:06 -07001514 BUG_ON(tsk->signal != current->signal);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001515 } while (tsk != current);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516 read_unlock(&tasklist_lock);
Oleg Nesterovf2cc3eb2008-02-08 04:19:06 -08001517
Linus Torvalds1da177e2005-04-16 15:20:36 -07001518 if (flag) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519 if (options & WNOHANG)
1520 goto end;
1521 retval = -ERESTARTSYS;
1522 if (signal_pending(current))
1523 goto end;
1524 schedule();
1525 goto repeat;
1526 }
1527 retval = -ECHILD;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528end:
1529 current->state = TASK_RUNNING;
1530 remove_wait_queue(&current->signal->wait_chldexit,&wait);
1531 if (infop) {
1532 if (retval > 0)
Oleg Nesterov9cbab812008-02-08 04:19:02 -08001533 retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001534 else {
1535 /*
1536 * For a WNOHANG return, clear out all the fields
1537 * we would set so the user can easily tell the
1538 * difference.
1539 */
1540 if (!retval)
1541 retval = put_user(0, &infop->si_signo);
1542 if (!retval)
1543 retval = put_user(0, &infop->si_errno);
1544 if (!retval)
1545 retval = put_user(0, &infop->si_code);
1546 if (!retval)
1547 retval = put_user(0, &infop->si_pid);
1548 if (!retval)
1549 retval = put_user(0, &infop->si_uid);
1550 if (!retval)
1551 retval = put_user(0, &infop->si_status);
1552 }
1553 }
1554 return retval;
1555}
1556
Eric W. Biederman161550d2008-02-08 04:19:14 -08001557asmlinkage long sys_waitid(int which, pid_t upid,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558 struct siginfo __user *infop, int options,
1559 struct rusage __user *ru)
1560{
Eric W. Biederman161550d2008-02-08 04:19:14 -08001561 struct pid *pid = NULL;
1562 enum pid_type type;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001563 long ret;
1564
1565 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1566 return -EINVAL;
1567 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1568 return -EINVAL;
1569
1570 switch (which) {
1571 case P_ALL:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001572 type = PIDTYPE_MAX;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573 break;
1574 case P_PID:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001575 type = PIDTYPE_PID;
1576 if (upid <= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577 return -EINVAL;
1578 break;
1579 case P_PGID:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001580 type = PIDTYPE_PGID;
1581 if (upid <= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583 break;
1584 default:
1585 return -EINVAL;
1586 }
1587
Eric W. Biederman161550d2008-02-08 04:19:14 -08001588 if (type < PIDTYPE_MAX)
1589 pid = find_get_pid(upid);
1590 ret = do_wait(type, pid, options, infop, NULL, ru);
1591 put_pid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001592
1593 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001594 asmlinkage_protect(5, ret, which, upid, infop, options, ru);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001595 return ret;
1596}
1597
Eric W. Biederman161550d2008-02-08 04:19:14 -08001598asmlinkage long sys_wait4(pid_t upid, int __user *stat_addr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001599 int options, struct rusage __user *ru)
1600{
Eric W. Biederman161550d2008-02-08 04:19:14 -08001601 struct pid *pid = NULL;
1602 enum pid_type type;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603 long ret;
1604
1605 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1606 __WNOTHREAD|__WCLONE|__WALL))
1607 return -EINVAL;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001608
1609 if (upid == -1)
1610 type = PIDTYPE_MAX;
1611 else if (upid < 0) {
1612 type = PIDTYPE_PGID;
1613 pid = find_get_pid(-upid);
1614 } else if (upid == 0) {
1615 type = PIDTYPE_PGID;
1616 pid = get_pid(task_pgrp(current));
1617 } else /* upid > 0 */ {
1618 type = PIDTYPE_PID;
1619 pid = find_get_pid(upid);
1620 }
1621
1622 ret = do_wait(type, pid, options | WEXITED, NULL, stat_addr, ru);
1623 put_pid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001624
1625 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001626 asmlinkage_protect(4, ret, upid, stat_addr, options, ru);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001627 return ret;
1628}
1629
1630#ifdef __ARCH_WANT_SYS_WAITPID
1631
1632/*
1633 * sys_waitpid() remains for compatibility. waitpid() should be
1634 * implemented by calling sys_wait4() from libc.a.
1635 */
1636asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
1637{
1638 return sys_wait4(pid, stat_addr, options, NULL);
1639}
1640
1641#endif